The Effect of Extreme Response and Nonextreme Response Styles on Measurement Invariance

• In Event: Response Styles and Measurement Invariance in Factor Modeling

Abstract

Extreme response style and non-extreme response style are prevalent in survey research using Likert-type scales. Their effects on measurement invariance (MI) in the context of confirmatory factor analysis (CFA) are systematically investigated through a Monte Carlo simulation study. More specifically, using the parameter estimates of a TIMSS data, we set up a population model. Two groups of original data and contaminated data with either one of the response styles are generated and analyzed via multi-group CFA with different constraints of MI. Preliminary results indicate the negative effects of response styles on MI has long been underestimated. The lack of MI between groups may be caused by the response styles, not the measured factors of focal interest.

Authors

• Min Liu, University of Hawaii - Manoa

• Allen G. Harbaugh, Murdoch University

• Jeffrey R. Harring, University of Maryland

• Gregory R. Hancock, University of Maryland